search

cnexah / DeepLineEncoding

Deep Line Encoding for Monocular 3D Object Detection and Depth Prediction
MIT License
15 stars 0 forks source link

What is bt and dht in 'yolomono3d_core.py'? #4

Closed sekimcb closed 1 year ago

sekimcb commented 1 year ago

Hello, again.

I wonder what do 'bt' and 'dht' mean in 'yolomono3d_core.py'.

And can I use hough transform idea as your model to MonoFlex model? If I could, is it ok use 'DLA' backbone to apply hough transform?

Thank you

cnexah commented 1 year ago

Thank you for your interest!

'bt' is the abbreviation for 'bottleneck', and is applied to reduce the number of feature channels. 'dht' is the abbreviation for 'deep Hough transform', and is applied to perform Hough transform on feature maps.

Yes, I think the idea of Hough transform can be applied to various models and backbones.

sekimcb commented 1 year ago

Thank you for your reply!! It helps me to understand!!

I have one more question... :) Where can I find "Line Pooling" module code in yours?

Thank you Sincerely

cnexah commented 1 year ago

The 'dht_backbone' in yolomono3d_core.py represents the 'Line Pooling' module. https://github.com/cnexah/DeepLineEncoding/blob/adbe91824480ebcd8cbb14f0cc766b465386b4b8/object_detection/visualDet3D/networks/detectors/yolomono3d_core.py#L51

sekimcb commented 1 year ago

Aha got it!

Thank you very much!